This project is developing an Internet architecture that enables users or their end-systems to select the paths their packets take through the network. Such user-controlled routes are desirable for both economic and technical reasons. Economically, user-controlled routes are a key factor in maintaining the competitiveness of the ISP marketplace, just as long distance carrier selection has had a lasting impact on telephony. Technically, user-controlled routes provide a fundamental means for improving the performance and reliability of network communications. This is because they allow end systems to use multiple diverse paths concurrently and reduce the dependence on a single network path that has undesirable characteristics. The key technical difficulties associated with user controlled routes lie in the areas of scalability and robustness. This project directly addresses these issues. It involves the development of the following components: 1) a scalable inter-domain routing protocol that distributes policy-constrained domain-level maps of the network to users and allows them to explicitly formulate domain-level routes; 2) an end system protocol stack that allows various applications to take advantage of path diversity; 3) algorithms and protocols that prevent users from rapidly switching paths in a fashion that jeopardizes the overall routing stability; 4) techniques that mitigate the security threats associated with user-controlled routes (e.g. source address spoofing, multi-path denial of service attacks).
If successful, this work will have a positive economic impact that fosters ISP competition, result in measurable technical improvements in terms of reliability and performance of network communications, and contribute to education. Algorithms, protocols, and software developed in this project will be disseminated through research publications and a project web-site.
